A data-driven crop model for maize yield prediction.

Area of Science:

• Agricultural Science
• Climate Change Adaptation
• Computational Modeling

Background:

• Accurate crop yield prediction is vital for global food security, especially with climate change impacts.
• Existing models often lack the integration of mechanistic understanding and data-driven computational power.
• Maize yield estimation is critical for agricultural planning and economic stability.

Purpose of the Study:

• To develop and validate a hybrid data-driven crop model for accurate maize grain yield prediction.
• To assess the model's performance using historical data from the US Corn Belt.
• To demonstrate the model's capability in explaining genotypic and environmental interactions and aiding seed selection.

Main Methods:

• A hybrid model combining process-based and data-driven approaches was developed.
• The model simulates daily biomass accumulation to estimate final grain yield.
• Computational experiments utilized extensive historical data (1981-2020) on maize yield, location, genotype, and environment.

Main Results:

• The proposed model achieved a 7.16% relative root-mean-square-error for average yield in 2020.
• The model provides scientifically explainable predictions, enhancing trust and understanding.
• It effectively identifies and differentiates interactions between genetic traits and environmental factors.

Conclusions:

• The hybrid model offers a significant advancement in accurate and explainable crop yield prediction.
• This approach holds potential for optimizing seed selection strategies to enhance farmer yields.
• The model contributes to more resilient agricultural systems in the face of environmental challenges.